A Practical Method for Short-term Earthquake Prediction Using Multiple High-frequency Tremor Events

The substantial damage induced by major earthquakes requires the preparation of prevention methods before serious shocks occur. For a half-century, researchers have tried to develop an efficient method for earthquake prediction using modern scientific methods without practical results except for two rare successes; as a result, the general evaluation is pessimistic. Among many phenomena, seismic activity has been the approach most often investigated. Particularly, foreshocks seem to offer the most potential. However, foreshocks are found to precede only a small fraction of major earthquakes and provide precursor parameters with too many kinds of diversity.

We need to find another seismic or similar phenomenon in the nucleation period with characteristics expected for foreshocks, i.e., a stable rate of occurrence and extremely large anomalies immediately before major earthquakes [1,2]. Here we make a special seismic catalog of high frequency tremors deduced anew from continuous seismic data of just before major earthquakes using the extensive network, High-net of Japan. Analyses of catalog of three major and one little bit smaller earthquakes show that there are three successive precursory phenomena, first at some six weeks, second at some four weeks, and finally immediately before the earthquake.

These results can provide evidence to predict major earthquakes without high rate of diversity among
threshold parameters; i.e., the selected threshold values for distinguishing precursor candidates are quite stable. Each precursory activity provides the three items of prediction, namely, occurrence time, epicenter and magnitude, with sufficient accuracy for practical disaster prevention efforts. The positive results will contribute to developing practical prediction methods to be used for the mitigation of serious earthquake disasters.

The proposed system is now in the level of POC, and expected to start in Japan without large difficulties because of sufficient level of observation network and storage of past data of some twenty years.